基于数据增广和CNN的地震随机噪声压制

卷积神经网络(Convolutional Neural Network,CNN)是一种基于数据驱动的学习算法,简化了传统从特征提取到分类的两阶段式处理任务,被广泛应用于计算机科学的各个领域.在标注数据不足的地震数据去噪领域,CNN的推广应用受到限制.针对这一问题,本文提出了一种基于数据生成和增广的地震数据CNN去噪框架.对于合成数据,本文对无噪地震数据添加不同方差的高斯噪声,增广后构成训练集,实现基于小样本的CNN训练.对于实际地震数据,由于无法获得真实的干净数据和噪声来生成训练样本集,本文提出一种直接从无标签实际有噪数据生成标签数据集的方法.在所提出的方法中,我们利用目前已有的去噪方法从实际地震数据中分别获得估计干净数据和估计噪声,前者与未知的干净数据具有相似纹理,后者与实际噪声具有相似的概率分布.人工合成数据和实际数据实验结果表明,相较于F-X反褶积,BM3D和自适应频域滤波算法,本文方法能更好地压制随机噪声和保护有效信号.最后,本文采用神经网络可视化方法对去噪CNN的机理进行了探索,一定程度上解释了网络每一层的学习内容.     

Seismic risk assessment considering cumulative damage due to aftershocks

Calculating the limit state (LS) exceedance probability for a structure considering the main seismic event and the triggered aftershocks (AS) is complicated both by the time‐dependent rate of aftershock occurrence and also by the cumulative damage caused by the sequence of events. Taking advantage of a methodology developed previously by the authors for post‐mainshock (MS) risk assessment, the LS probability due to a sequence of mainshock and the triggered aftershocks is calculated for a given aftershock forecasting time window. The proposed formulation takes into account both the time‐dependent rate of aftershock occurrence and also the damage accumulation due to the triggered aftershocks. It is demonstrated that an existing reinforced concrete moment‐resisting frame with infills subjected to the main event and the triggered sequence exceeds the near‐collapse LS. On the other hand, the structure does not reach the onset of near‐collapse LS when the effect of triggered aftershocks is not considered. It is shown, based on simplifying assumptions, that the derived formulation yields asymptotically to the same Poisson‐type functional form used when the cumulative damage is not being considered. This leads to a range of approximate solutions by substituting the fragilities calculated for intact, MS‐damaged, and MS‐plus‐one‐AS‐damaged structures in the asymptotic simplified formulation. The latter two approximate solutions provide good agreement with the derived formulation. Even when the fragility of intact structure is employed, the approximate solution (considering only the time‐dependent rate of aftershock occurrence) leads to higher risk estimates compared with those obtained based on only the mainshock. Copyright © 2016 John Wiley & Sons, Ltd.     

Effects of sand content on initial gravel motion in gravel‐bed rivers

When fine sediments are present in gravel streambeds (gravel‐framework beds), the gravel can be more easily removed from its original position, compared with gravel in a streambed without fine sediment but otherwise under the same hydraulic conditions. In this study, the effect of the presence of sand on the initiation of gravel motion in gravel riverbeds was investigated using flume experiments. The relationship between the critical Shields stress for gravel motion initiation and the fraction of sand in the bed was determined experimentally. The results can be summarized as follows. (1) When the fraction of sand in the bed is smaller than about 0.4, the critical Shields stress for the initiation of gravel motion decreases with increasing fraction of sand. The critical Shields stress increases, however, with increasing fraction of sand when it is larger than about 0.4. (2) The difference between the value of the critical Shields stress predicted by the Egiazaroff equation and the value obtained from the experimental data becomes maximum at about 0.4 of the fraction of sand. Here an empirical relation between the critical Shields stress and the fraction of sand is proposed so as to consider the effects of the ratio of the characteristic gravel size to the mean size of the bed material on the critical Shields stress. (3) Gravel in armored beds can be more easily mobilized by supplying sand as part of a sediment augmentation scheme. The sand fraction in the subsurface layer of the bed appears to reduce the friction angle of exposed particles. Sediment augmentation using sand has been recently demonstrated to be a viable alternative for mobilizing gravel for the restoration of gravel‐bed rivers downstream of dams. The quantitative evaluation obtained through the experiments reported here may be useful for the design of augmentation schemes. Copyright © 2017 John Wiley & Sons, Ltd.     

循环荷载下加筋砾性土填料的动三轴试验分析

为探究循环动载作用下加筋砾性土填料的动力特性,在不同加筋层数和围压下对加筋砾性土进行了固结不排水动三轴试验,研究加筋层数和围压对加筋砾性土动力特性的影响,并进一步分析了加筋砾性土轴向累积应变发展机制。研究表明:加筋层数增加时,轴向累积应变减小,回弹模量增大,且加筋作用的影响幅度逐渐衰减;增大围压时,土体轴向累积应变减小,回弹模量和动孔压均随之增大;随着加筋层数和振次的增加,滞回曲线逐渐向应力轴靠近,滞回圈面积逐渐减小,土体耗能作用减弱。基于安定理论和间接影响带理论,揭示了加筋作用对轴向累积应变发展的影响机制。建立了能够反映加筋层数的加筋砾性土轴向累积应变预测模型,其条件参数a、b、g与加筋层数呈线性关系,可有效预测循环荷载下加筋砾性土路基沉降变形规律。     

Shoreline evolution of the Holland coast on a decadal scale

The decadal evolution of the dunefoot position along 120 km of the Holland coast is analysed. Firstly, a new definition of dunefoot is introduced based on hydrodynamic and morphologic considerations, which is not affected by local and instantaneous processes occurring immediately before the bathymetric survey. The dunefoot evolution over decadal scales indicates the existence of spatial and temporal oscillations in the shoreline position with magnitudes of 2–3 km length and a periodicity of 4–15 years. Two main controlling factors of this behaviour are identified: (1) influence of sub-aqueous bar systems, and (2) changes in the storm-wave conditions reaching the coast. Although the precise controlling processes of the relation between the dunefoot and the subaqueous profile still remain unclarified, we introduce the concept that the development of a morphological bar cycle requires a fixed amount of time-integrated forcing that is proportional to the cumulative effect of storm waves. Beach mobility along the Holland coast on decadal scales (10–20 m) is similar or lower than mobility introduced by storms or by seasonal cycles. However, it is important to consider these changes for the possible implications on the local vulnerability of the coast to extreme events. Copyright © 1999 John Wiley & Sons, Ltd.     

COMPARISON STUDY OF TWO KINDS OF CODES TO MEASURE FAULT-OFFSETS BASED ON MATLAB: A CASE STUDY ON EASTERN ALTYN TAGH FAULT

Geomorphic offsets displaced by coseismic surface rupture can be analyzed to identify earthquake recurrence behavior. Therefore, obtaining a sufficient and precise along-fault offset dataset is vital to identify long-term earthquake recurrence behavior. Furthermore, knowledge of along-fault slip distribution during a single-earthquake or multi-earthquakes is important for other reasons, including a better understanding of the relationship between earthquake size and coseismic displacements, fault kinematics and fault mechanics. A recent flourish of offsets-measuring software and high-resolution topographic data together offer an unprecedented opportunity to measure high-density fault offsets. Here, we introduce and compare two kinds of most popular software, LaDiCaoz and 3D_Fault_Offsets. We describe the workflow and principle of the two codes by taking a fault-offset example on the eastern Altyn Tagh Fault. LaDiCaoz iterates over the channel morphology and position parameters and determines the summed absolute elevation difference Σ[Δ(elevation)] between both transverse profiles. The optimal horizontal offset is defined by the parameter combination that results in the least mismatch between two profiles. Compared with LaDiCaoz, the principle of 3D_Fault_Offsets is more complicated by measuring the offset in three dimensions. It mathematically identifies and represents nine of the most prominent geometric characteristics of common sublinear markers along faults in three dimensions, such as the streambed(minimum elevation), top, free face and base of channel banks or scarps(minimum Laplacian, maximum gradient, and maximum Laplacian), and ridges(maximum elevation). By calculating best fit lines through the nine point clouds on either side of the fault, the code computes the lateral and vertical offsets between the piercing points of these lines onto the fault plane, providing nine lateral and nine vertical offset measures per marker. Through a Monte Carlo approach, the code calculates the total uncertainty on each offset. Although both 3D_Fault_Offsets and LaDiCaoz are developed based on the Matlab platform, there are significant differences in principles, linear marker, software interface, repeatability, input-file types, degree of automation, adaptability, output file types, etc. In this part, we compare and summarize their features, advantages, and disadvantages. Finally, we calculate the correlation of two groups of fault-offset data derived from the two methods along the eastern ATF. By doing this, we try to explore if the two methods can be crosschecked and to study how sinuosity of the linear geomorphic markers affect the measuring results. By discussing and comparing the accuracy of the two measuring methods, we consider that LaDiCaoz is better than 3D_Fault_Offsets in accuracy aspect. In our opinion, there exist some disadvantages in the both software, and higher automation and introduction of artificial intelligence will be the future development direction.     

北斗广域增强系统星钟和星历误差改正算法研究

区域北斗星基增强系统提供等效钟差改正数统一修正星历和钟差误差。随着系统的建设发展,新一代北斗星基增强系统将区分星历和钟差误差改正信息,以提高差分改正精度。由于北斗卫星混合星座设计及区域监测网的局限,星历和钟差误差的高精度分离计算面临着新的挑战。对北斗星基增强系统的星历和钟差改正算法进行了研究,分别采用动力学和运动学模式计算了卫星星历和钟差改正数,并基于北斗实测数据,对两种处理模式的差分改正精度进行了对比研究。试验结果表明,采用动力学和运动学差分方法,得到的双频伪距实时定位精度分别为1.76m和1.78m,定位精度与WAAS及EGNOS相当。利用运动学和动力学差分改正数后均可得到分米级的精密单点定位(precise point position,PPP)结果,其中采用动力学广域差分改正数,收敛后定位精度可达到15cm;采用运动学广域差分改正数,收敛后定位精度可达45cm。     

北斗广域差分分米级定位的分区切换算法

从2017年1月开始,北斗系统在播发基本导航电文的同时,提供了包含实时轨道改正数、钟差改正数、电离层格网改正数和分区综合改正数四重广域差分参数,使得系统具备了基于系统广播的电文参数实现广域单站实时动态分米级定位的能力。实际应用表明,北斗系统播发的分区综合改正数存在中断的情况,使得实时定位存在因重新收敛而造成跳变的情况。本文分析了分区中断期间定位跳变产生的原因,提出一种适用于分区切换的广域差分分米级定位算法。采用中国境内7个北斗测站10d的静态数据和车载动态数据对新算法进行评估验证。结果表明,在分区改正数中断的情况下,分区切换算法得到的坐标误差与使用连续分区综合改正数结果一致,双频动态定位水平和高程方向分别小于0.3m和0.5m,平均三维定位精度优于0.5m;车载实时动态定位测试中,分区切换前后的单双频定位精度不受切换影响,与切换前保持一致。分区切换算法有效保证了分区改正数中断后北斗广域分米级实时动态定位的连续性。     

动态GNSS监测瞬态无震蠕滑信号提取

断层的瞬态无震蠕滑常诱发震级高、破坏性强的地震。针对其滑动速度缓慢,难以探测的问题,本文基于GNSS连续坐标时间序列的异常波动提出一种断层瞬态无震蠕滑信息的自动探测方法。首先利用独立成分分析进行时空滤波,提高坐标时间序列的信噪比;然后计算坐标时间序列波动的相对强度指数及峰度值;最后通过累积分布函数将其转换为蠕滑信号概率,进而探测断层蠕滑事件。本文模拟500 d GNSS地表位移时间序列进行仿真试验,其中包含25 d瞬态蠕滑信号。试验结果表明,当信号强度至少与噪声水平相当时可有效探测出断层的蠕滑信息。计算Akutan地区连续3年的GNSS数据后探测到一个蠕滑信号,推断其可能为火山岩强烈运动引起的无震蠕滑。通过对四川省陆态网18个测站7年的GNSS数据处理后发现了4个异常信号。     

非差增强信息与差分虚拟观测数据的等价变换理论

针对目前非差精密单点定位增强信息无法直接用于RTK(real time kinematic)相对定位的问题,研究了基于附加坐标约束的参考站非差精密单点模糊度固定解提取非差改正信息的方法,并建立了非差增强信息与虚拟参考站观测信息等价变换模型,重点论述了空间状态域信息(state space representation,SSR)在等价变换中的区别应用。根据RTK模糊度部分固定技术,利用实测数据设计实验证明了算法的正确性与可用性。结果表明,虚拟零基线可获得与网络RTK同等精度的定位效果,从而实现了区域增强系统在非差与差分模式上的高度统一。